PolarDB:2024 compute node release notes

Category

Description

New features & improvements

Added support for parallel SortWindow execution optimization.

Added support for the INFORMATION_SCHEMA.INNODB_SYS_TABLESPACES view.

Optimized the balance of task allocation by the DDL MPP scheduler across multiple nodes.

Improved the performance of partition routing calculations for a Clustered Columnar Index (CCI).

Optimized the metadata cleanup process for DRDS mode databases to avoid deadlocks.

Optimized the result display of the SELECT polardb_version(); statement.

Optimized the process of deleting metadata in the columnstore engine to reduce GMS resource consumption.

Optimized the logic for executing add partition operations on partitioned tables.

Fixed issues

Fixed an issue where the SHOW DBSTATUS, SHOW TABLEINFO, and REBALANCE statements did not retrieve real-time statistics for tables with MySQL 8.0 as the storage engine.

Fixed a failure in TRUNCATE or DROP partition operations on tables without a Global Secondary Index (GSI) or CCI. The failure was caused by the presence of a GSI or CCI on other tables in the same table group.

Fixed a Baseline deletion failure caused by a simultaneous write request.

Fixed an issue with collecting statistics for columns named after HyperLogLog (HLL) reserved words.

Fixed an issue where lock-free column type changes caused global clustered index tables to lose some Local Indexes.

Fixed an issue where repeated serialization of statistics histograms caused space usage to expand.

Fixed an issue where Baseline Hint Bind became invalid due to table schema changes.

Fixed an issue where creating unnamed full-text indexes on foreign key tables failed.

Fixed an issue where metadata read errors occurred when executing physical DDL on logical tables with many shards.

Fixed an issue where vectorized IN operations on string columns failed when NULL values were present.

Fixed a connection leak issue when the Follower copied database data file snapshots during physical replication.

Category

Description

New features & improvements

Added support for collecting HLL (HyperLogLog, a probabilistic algorithm) related statistics for databases and tables in parallel across storage nodes.

Added support for lock-free schema changes through DMS.

Columnstore indexes now support broadcasting the left table in SEMI HASH JOIN.

Optimized the optimizer's constant inference and JOIN pushdown logic.

Optimized issues with service traffic interruptions and jitter during maintenance operations on underlying storage nodes.

Optimized the metadata cleanup process for DRDS mode databases to avoid deadlocks.

Optimized the result display of the SELECT polardb_version() statement.

Optimized the process of deleting metadata in the columnstore engine to reduce GMS resource consumption.

Optimized the performance of columnstore indexes when computing route partitions.

Optimized the execution plan for columnstore indexes during partition pruning.

Optimized the execution logic for adding partitions to partitioned tables.

Fixed issues

Fixed an issue where the SHOW DB STATUS, SHOW TABLE INFO, and REBALANCE statements did not retrieve real-time statistics for tables with MySQL 8.0 as the storage engine.

Fixed an issue where executing the CREATE DATABASE LIKE statement caused a system error when the COLLATE property of the source database was empty.

Fixed an issue where the SHOW DDL statement displayed inaccurate data backfill progress.

Fixed an issue where columnstore index distinct agg results might be incorrect when ENABLE_EXPAND_DISTINCTAGG was turned off.

Fixed an issue where scheduled tasks related to HLL (HyperLogLog, a probabilistic algorithm) could not be initiated normally.

Fixed an issue where executing SHOW CREATE TABLE on tables with localtity markers in columnstore read-only instances resulted in errors.

Fixed an issue where creating unnamed full-text indexes on foreign key tables failed.

Fixed a connection leak issue when the Follower copied data file snapshots during physical replication.

Fixed an issue where using RangeScan when executing ORDER BY ASC statements resulted in unexpected sorting.

Fixed an execution failure when vectorizing an IN clause on a string column that contained NULL values.

Fixed an error when reading metadata from a logical table with a large amount of data.

Fixed an issue where Direct HINT did not correctly replace physical table names.

Fixed an invalid group or physical table names error that could occur during a query on a table undergoing partition migration.

Fixed an illegal routing calculation error caused by value truncation in an equality query on a prefix partition column.

Category

Description

New features & improvements

Tables with columnstore indexes now support TRUNCATE TABLE operations. For more information, see Columnstore index DDL statement limitations.

During scale-in and scale-out operations, the system now automatically identifies partitioned tables without GSI and performs partition pruning on queries issued by UPSERT statements based on the primary table. For more information, see Global secondary index (GSI).

Optimized the metadata deletion process in columnstore read-only instances to reduce Metadata Center (GMS) resource consumption.

Optimized the order of scanning physical partitions when executing UPDATE ORDER BY LIMIT statements to reduce the number of partitions scanned.

Fixed issues

Fixed an issue where rollback failed if the connection was interrupted during the execution of an ALTER TABLE MODIFY statement.

Fixed an issue where the system reported an error when executing the CREATE DATABASE LIKE statement when the COLLATE attribute in the source database was an empty string.

Fixed an issue where temporary table memory was not released after subquery iteration was completed.

Fixed an issue where FORCE INDEX did not support specifying columnstore archived tables. For more information, see FORCE INDEX.

Fixed an issue where table-specific HINTs were not effective in some queries containing IN clauses. For more information, see HINT.

Fixed an issue where executing ORDER BY ASC statements using RangeScan resulted in unexpected sorting.

Fixed an issue where querying during partition migration could occasionally result in invalid group or physical table names errors.

Fixed an issue where the system reported errors when reading metadata for logical tables with massive data.

Fixed an issue where the BKAJoin and SortMergeJoin operator combination resulted in incomplete result sets when there were many matching results in the join.

Fixed an issue where deadlocks could occur when database deletion operations coincided with scale-in or scale-out operations.

Fixed an issue where ProjectWindowTransposeRule caused Window queries to report errors.

Fixed an issue where Direct HINT did not correctly replace physical table names.

Fixed an issue where equality queries on prefix partition columns resulted in illegal route calculation errors due to query value truncation.

Fixed an issue where the PARSER specified for full-text indexes was lost when creating new tables.

Fixed an issue where ProjectToWindowRule caused optimizer errors.

Fixed an issue where SemiJoinJoinTransposeRule caused incorrect execution results.

Fixed an issue where SQL execution failed due to global index selection on tables with DECIMAL primary keys.

Fixed an issue where the BKAJoin operator failed during execution due to type inconsistencies.

Fixed an issue where scheduled tasks related to HLL (HyperLogLog, a probabilistic algorithm) could not be initiated normally.

Fixed an issue where calculating Selectivity in subquery execution plans caused errors, resulting in query failures.

Fixed an issue where the Spill counter was not reset to zero after temporary tables were written to disk.

Fixed a null pointer issue caused by ProjectWindowTransposeRule.

Category

Description

New features & improvements

During scale-in and scale-out operations, the system now automatically identifies partitioned tables without GSI and performs partition pruning on queries issued by UPSERT statements based on the primary table. For more information, see Global secondary index (GSI).

When closing columnstore nodes, the system now automatically cleans up the columnstore Plan Cache.

Tables with columnstore indexes now support TRUNCATE TABLE operations. For more information, see Columnstore index DDL statement limitations.

Optimized GSI construction performance (concurrent backfilling by globally ordered primary key ranges during construction). For more information, see Global secondary index (GSI).

Optimized SPM (SQL Plan Management) bound columnstore plan selection to be unaffected by the global columnstore routing switch.

Optimized Metric View to display system Metric information.

Optimized resource limitations and scheduling efficiency for logical and physical Backfill operations.

Optimized the HLL (HyperLogLog, a probabilistic algorithm) collection logic to avoid scanning unique keys.

Optimized the performance overhead of MDL lock wait detection. For more information, see MDL optimization (metadata lock).

Optimized the columnstore index selection logic (SQL containing Apply operators is no longer routed to columnstore nodes). For more information, see Row-column routing mechanism.

Optimized the workload type determination logic, configuring columnstore query workload type as AP type by default. For more information, see HTAP mixed workload.

Optimized the performance of the SHOW DDL statement when viewing complex DDL.

Optimized the health check interval and timeout parameters of the high availability (HA) system to reduce recovery time.

Optimized the granularity settings of scale-in/scale-out DDL Tasks to improve scheduling efficiency.

Fixed issues

Fixed an issue where metadata errors could occur when executing multiple TRUNCATE operations on the same table in DRDS mode databases with the table recycle bin feature enabled. For more information, see AUTO mode database and DRDS mode database.

Fixed an issue where regular privilege accounts on newly created instances would directly report errors instead of returning warnings when executing SET GLOBAL statements.

Fixed an issue where single-value frequencies in statistics were estimated as 0 in histograms.

Fixed an issue where TTL definitions were lost when CREATE TABLE WITH TTL statements were transmitted to downstream Replicas through CDC. For more information, see TTL.

Fixed an issue where the columnstore optimizer lost Sort operation results, causing unexpected results.

Fixed an issue where hidden primary keys became unavailable when executing TRUNCATE in RECYCLEBIN mode.

Fixed an issue where SAMPLE RATE expressed in scientific notation caused storage node SAMPLE HINT to be ineffective.

Fixed an issue where the system incorrectly matched execution plans for multiple IN parameters during preprocessing when SQL statements contained IN clauses with only one parameter.

Fixed issues where the SQL parser failed to parse ALTER statements containing implicit table groups, and marking statements could not be processed correctly. For more information, see Table group types.

Fixed an issue where FORCE INDEX CCI was ineffective on single tables (tables without sharding). For more information, see FORCE INDEX.

Fixed an issue where SHOW DDL STATUS displayed incorrect rates during concurrent backfilling.

Category

Description

New features & improvements

Primary instances now support executing columnstore index-related DDL statements (this feature can be configured with a switch, enabled by default). For more information, see Configure columnstore index switches.

SPM (SQL Plan Management) now supports binding DN HINT.

Columnstore indexes now support partition splitting.

Fixed issues

Fixed an issue with duplicate Feedback information in statistics.

Fixed an issue where single-value frequencies in statistics were estimated as 0 in histograms.

Fixed an issue where newly created regular privilege accounts on newly created instances would directly report errors instead of returning warnings when executing SET GLOBAL statements.

Fixed an issue where validation errors occurred when column types were changed in lock-free mode.

Fixed an issue where PARTIAL_SELECTION was processed incorrectly when a single Chunk contained NULL values during Selection generation, leading to incorrect results.

Fixed an issue where FORCE INDEX CCI was ineffective on single tables.

Fixed compatibility issues with certain SHOW statements (SHOW INDEX FROM, SHOW TABLE STATUS, DESC TABLE/SHOW COLUMNS FROM, SHOW COLLATION, SHOW CHARACTER SET, SHOW PRIVILEGES) in columnstore read-only instances.

Fixed an issue where incorrect results were output due to filter conditions being erroneously passed through to outer joins (OUTER JOIN) in certain scenarios.

Fixed an issue where executing INSERT IGNORE during subpartition changes could occasionally result in null pointer errors.

Fixed an issue where NDV (number of distinct values) collection mistakenly deleted column statistics during table topology changes.

Fixed an issue where the thread building the snapshot was blocked when CDC was building a full Schema snapshot.

Fixed an issue where automatic FORCE INDEX for indexes containing reserved keywords failed due to missing backticks (`), causing SQL errors.

Fixed an issue where partition key pruning after using Lookup Join (BKAJoin) optimization technique on statements containing Null-Safe equals operators (<=>) resulted in query column disorder, causing incorrect results. For more information, see Lookup Join (BKAJoin).

Category

Description

New features

Added support for reusing execution plans regardless of whether SQL statements end with a semicolon.

Added support for automatically routing single-table query statements with columnstore indexes to read-only instances.

Added support for prohibiting built-in database names such as MySQL when creating databases.

Added support for subqueries in JOIN ON conditions during single-table pushdown.

Added support for DDL operations on primary tables with columnstore indexes by default.

Added default support for using columnstore indexes with ANALYZE TABLE.

Added support for creating function indexes on JSON type fields when the engine version is MySQL 8.0.

Fixed issues

Fixed an issue where DML pushdown markers remained when changing column types in lock-free mode.

Fixed an issue where DDL remained in execution state and could not complete when executing ALTER TABLE to modify column names if the case of the new column name did not match the original.

Fixed an issue where data in the INFORMATION_SCHEMA.TABLE could not be updated in real-time when the DN node engine version was MySQL 8.0.

Fixed an issue where querying the INFORMATION_SCHEMA.STATISTICS table in PREPARE mode resulted in errors.

Fixed an issue where columnstore index table creation failed due to the default table group being reassigned during partition changes.

Fixed an issue where filtering aggregation results without grouping could produce unexpected results.

Fixed an issue where columnstore archived table creation failed when time type column values in TTL tables were zero or empty.

Fixed issues where the SQL parser failed to parse ALTER statements containing implicit table groups, and marking statements could not be processed correctly.

Fixed an issue where CDC SQL generated when adding CCI partitions to TTL tables contained random suffixes, causing CDC downstream synchronization exceptions.

Fixed an issue where DirectPlan pushdown did not work as expected for cross-database single-table INSERT SELECT operations.

Category

Description

Fixes

Fixed an issue where database users could not be created or queried through Navicat.

Fixed an issue where DataGrip continuously displayed Warning messages when connecting to the database.

Fixed an issue where database views could not be queried through DBeaver.

Category

Description

Optimization

Optimized the scale-in/scale-out process to check if the source table exists during rollback.

Resolution

Fixed an issue where DML pushdown markers remained when changing column types in lock-free mode.

Fixed an issue where SQL with NODE HINT had the same template ID as other SQL.

Fixed an issue where validation might fail when adding NOT NULL constraints while changing column types in lock-free mode.

Fixed an issue where filtering aggregation results without grouping could produce unexpected results.

Fixed an issue where special characters caused statistics histogram serialization failures.

Fixed an issue where querying the INFORMATION_SCHEMA.STATISTICS table in PREPARE mode resulted in errors.

Fixed an issue where DDL remained in execution state and could not complete when executing ALTER TABLE to modify column declarations if the case of the new column name did not match the original.

Fixed an issue where STR_TO_DATE('2020-08-01 00:00:00', '%Y-%m-%d %T') produced unexpected results.

Category

Description

New features & improvements

Added SPM (SQL Plan Management) support for columnstore read-only instances.

Added support for automatically adding or removing quotes for parameter values with mismatched types.

Added support for character type SortKeys in the columnstore read-only instance index pruning feature.

Added support for automatically scheduling columnstore queries from primary instances to execute on columnstore read-only instances.

Optimized the execution logic of aggregate (AGG) functions: in row storage scenarios, pushdown aggregation is used by default; in column storage scenarios, two-stage aggregation is used by default.

Optimized the index selection logic for hot spot values.

Optimized the performance of NOT IN expressions in columnstore read-only instance queries.

Fixed issues

Fixed an issue where the left operand in WHERE clauses of SQL statements only supported integers in the range [-2⁶³,2⁶³-1].

Fixed an issue where Baseline did not support union statements.

Fixed an issue where SPM (SQL Plan Management) execution plan concurrency calculation was unexpected in MPP (Massively Parallel Processing) mode.

Fixed a security issue with the AES_DECRYPT function in high-concurrency environments.

Fixed an issue where special characters caused statistics histogram serialization failures.

Fixed an issue where execution plans lost the columnstore attribute distribution during deserialization.

Fixed an issue where change operations failed on sequence with names longer than 64 characters.

Fixed an issue where different alias case caused template ID inconsistency when using aggregate (AGG) functions.

Fixed an issue where CREATE DATABASE statements could create mysql, performance_schema type databases.

Fixed an issue where executing SQL with LIMIT in columnstore read-only instances could occasionally produce unexpected results.

Fixed an issue with column mismatches when updating child tables with foreign key constraints.

Fixed an issue where SQL containing aggregate functions with subqueries reported errors during execution.

Fixed an issue where validation might fail when changing column types and adding NOT NULL constraints in lock-free mode.

Fixed an issue where join type inference failed when pushing down predicate conditions into subqueries.

Fixed an issue where column matching failed in INSERT SELECT statements containing aggregate (AGG) functions.

Modification

Changed the default value of the switch controlling whether primary tables with columnstore indexes can undergo DDL operations to enabled.

Category

Description

Fixed issue

Fixed a memory overflow issue caused by excessively large cache pool maximum memory configuration.

Category

Description

New features & improvements

Added Plan Cache (execution plan cache) support for columnstore read-only instances.

Added support for scheduling columnstore queries to execute on columnstore read-only instances.

Optimized the performance of SQL containing JOIN keywords.

Optimize the performance of UPDATE statements that contain ORDER BY LIMIT clauses when they are executed on partitioned tables.

Optimized the display strategy for transaction-related variables.

Optimized the distributed deadlock detection mechanism.

Fixed issues

Fixed an issue where executing IN(NULL) queries on indexes containing string type columns resulted in errors.

Fixed an issue where the TO_BASE64() function produced incorrect encoding results when applied to BLOB fields.

Fixed an issue where SQL optimized by JoinAggToJoinAggSemiJoinRule produced incorrect results.

Category

Description

New features & improvements

Added functionality to reuse execution plans for SQL statements regardless of whether they end with a semicolon.

Added support for subqueries in JOIN ON conditions during single-table pushdown.

Optimized cardinality estimation in extreme data skew situations.

Optimized the scale-in/scale-out process to check if the source table exists during rollback.

Fixed issues

Fixed an issue where SAMPLE HINT was ineffective when SAMPLE RATE was expressed using scientific notation.

Fixed an issue with routing and value changes when inserting data using the TIMESTAMP() function.

Fixed an issue where the left operand in WHERE clauses of SQL statements only supported integers in the range [-2⁶³, 2⁶³-1].

Fixed an issue with unexpected DDL rollbacks.

Category

Description

New features & improvements

Added automatic force index functionality for certain scenarios.

Added support for specifying precision for float and double data types.

Added rate limiting support for columnstore read-only and read-only instances.

Optimized Information_schema.tables data display.

Optimized concurrent limitations for physical execution of ALTER TABLE related DDL.

Increased the rate limiting thresholds for wait and concurrent on DN nodes.

Fixed issues

Fixed an issue where IN pruning took a long time when the number of pushed-down tables exceeded 9.

Fixed an issue where lock-free column type changes occasionally failed when converting string types to non-string types.

Fixed an issue where baseline delete commands took effect on some nodes but not on others.

Fixed an issue where adding auto_increment to fields in tables without sequences using lock-free column type changes caused data insertion errors.

Fixed an issue where isolation level strings could be decoded incorrectly due to case sensitivity and underscore problems.

Fixed an issue where baseline add/fix commands calculated inaccurate costs for execution plans with IN parameters.

Category

Description

New features & improvements

Added support for reading snapshot data using AS OF TSO syntax.

Optimized Distinct Agg query plans.

Fixes

Fixed an issue where CDC marking actions were missing when recreating a database after a DROP DATABASE failure, causing errors.

Category

Description

New features & improvements

Added support for the INSERT INTO TABLENAME() VALUES() syntax.

Added support for using tableless UNION syntax in UPDATE SET clauses.

Optimized logical REPLACE pushdown judgment to support pushing down REPLACE statements in some scenarios under RC isolation level.

Prohibited executing DDL without specifying SCHEMA in the INFORMATION_SCHEMA database.

Fixed issues

Fixed an issue where single-table sample hints for statistics were ineffective.

Fixed an issue where OMC 2.0 could cause incorrect routing of UGSI historical data.

Fixed an issue where foreign key constraints failed to be added due to issues with self-referencing table index and character set detection.

Fixed an issue where LIMIT x, y parsing failed when values were too large.

Category

Description

New features & improvements

Added physical replication support for encrypted tables.

Added support for scale-in/scale-out maintenance windows.

Added support for specifying KEY_BLOCK_SIZE and other table compression-related parameter configurations when creating tables.

Optimized the execution speed of GSI creation and table partition key changes.

Fixed issues

Fixed an issue where migrating UGSI reported errors when physical replication was enabled.

Fixed an issue where lock-free column changes on single tables and broadcast tables failed to execute.

Category

Description

New features and improvements

Added support for MySQL-compatible multi-statement modifications in alter table.

Optimized index selection logic to reject GSIs that cannot use local indexes for DML operations involving only one table.

Optimized by removing redundant group by uk.

Optimized online column type change execution speed and compute node CPU usage.

Optimized the performance of IN expression statistics estimation in IN Values scenarios.

Fixed issues

Disabled Parser from parsing private DDL in comments in global Binlog Query Events to prevent downstream DDL application errors.

Fixed an issue where MDL lock cleanup judgment errors during DDL execution caused transaction draining to fail, leading to data inconsistency.

Fixed an issue where insert ignore could report Duplicate entry errors during scale-in/scale-out operations.

Fixed an issue where result sets were not properly closed in MPP mode, causing backend connections to hang.

Fixed an issue where inspect index gave incorrect recommendations in certain scenarios.

Fixed an issue where inner join-outer join reorder could cause incorrect results.

Fixed an issue where DDL could unexpectedly fail when auto_increment columns contained records with value 0.

Fixed an issue where metadata import failed during Standard Edition to Enterprise Edition upgrades when database names ended with s or S.

Fixed an issue where alter table remove partitioning with specified options reported DDL execution errors.

Fixed an issue where storage node port resource exchange in low-resource scenarios caused misuse of private protocol instance-level connection pools.

Disabled XA_TSO transaction strategy in hot spot update scenarios to avoid transaction type incompatibility issues when upgrading instances.

Category

Description

Fixed issues

Fixed an issue where column default value types were incorrect after changing the Default expression of columns.

Fixed an issue where route calculation was incorrect for string type negative integers (such as '-184467440737095516150') with unsigned type partition keys in MySQL 8.0 mode.

Category

Description

New features and improvements

Added support for using columnstore tables with analyze table to accelerate statistics collection.

Added the ability to isolate execution plans by instance.

Optimized long string statistics.

Optimized the TopN construction logic for statistics.

Unified the plancache expiration time for read-only/primary instances to 12 hours by default.

Optimized the version number prefix returned by select version(), upgrading from 5.6.29 to 5.7.25.

Fixed issues

Fixed an issue where using count(1) in window functions caused errors.

Fixed an issue with incorrect cardinality estimation after constant folding.

Fixed an issue where hex/length functions produced incorrect calculation results for blob types.

Fixed an issue where statistics collection failed when tables contained columns named ```x```.

Fixed an issue where count(*) could not perform column pruning on window functions.

Fixed an issue where the show index from statement displayed physical table names.

Fixed an issue where the whitelist manager did not interrupt promptly when CN initialization reported errors while loading metadata with low probability, causing whitelist metadata loading errors.

Fixed a counting error in text Prepare protocol Deallocate.

Fixed an issue where creating ddl seq would clear the entire schema's baseline.

Fixed an issue where CBO might not be able to push down agg to GSI.

Category

Description

New features & optimizations

Added Online Modify Column 2.0 functionality.

Added binlog-related maintenance commands such as show full binary logs and show full master status.

Optimized the creation and deletion process of stored procedures to avoid metadata residue issues.

Reduced CPU consumption when executing REPLACE/UPSERT statements on tables containing GSI or UGSI.

Fixed issues

Fixed an issue where some old execution plans were not cleaned up when creating local indexes on tables containing clustered indexes.

Fixed an issue where physical tables could not be found when executing multi-table RENAME in DRDS mode databases.

Fixed an issue where concurrent scale-in/scale-out operations on different storage pools could cause incorrect CDC DN list calculations.

Fixed an issue with DML execution errors in foreign key tables.

Fixed an issue where distinct count on multiple columns could report errors.

Fixed an issue where the statistics correction feature was ineffective due to incorrect case settings.

Fixed an issue where drop table statements containing multiple database names could still execute successfully.

Fixed an issue where creating tables in AUTO mode databases reported errors when partition columns of partitioned tables contained backtick characters.

Fixed an issue where insert ignore (returning) could cause data validation failures during scale-in/scale-out or UGSI creation.

Fixed an issue with duplicate index names between GSI and Local Index.

Fixed an issue where UGSI creation failed if the index name contained special characters.

Fixed an issue with Row expression calculation errors.

Fixed an issue where transaction log cleaning frequency was too fast for non-UTC+8 instances.

Fixed issues with multi-stage DDL routing errors on read-only instances, SQL_MODE processing causing modify column not null to hang, and DDL execution hanging on logical tables with special characters in table names.

Category

Description

Optimization

Optimized sharding pushdown scenarios to enable xplan cache hits, accelerating high-concurrency performance.

Fixed issues

Fixed an issue where certain views in INFORMATION_SCHEMA could not be queried normally with IN query conditions.

Fixed a physical SQL syntax issue with BKA Join in special character scenarios.

Fixed an issue where association matching was disordered when subqueries had three or more association items.

Fixed an issue with incorrect route calculation in BKA Join scenarios.

Category

Description

New features & optimizations

Added the show create tablegroup command to support displaying predefined table groups.

Added atomic commit across multiple shards for DDL operations pushed down to DN, including adding/removing columns, adding/removing indexes, and modifying columns.

Added display of specific failure reasons when ANALYZE TABLE fails.

Added streaming result set retrieval for sampling.

Optimized subquery decorrelation logic to ignore the impact of order by without limit in subqueries on decorrelation.

Optimized online column type change operations to reduce change time.

Fixed issues

Fixed support for global configuration of max_allowed_packet and max_user_connections.

Fixed an issue where automatically sharded single tables could not directly modify primary keys.

Fixed an issue where querying timestamp fields in archived tables occasionally produced incorrect results.

Fixed an issue where physical shards could not be found after the rename table command was interrupted.

Fixed an issue where inspect index could not diagnose when the current database contained views.

Fixed an issue where DDL could unexpectedly pause when modifying the column type of global unique index shard key columns.

Fixed an issue where TTL tables occasionally failed when executing the rename table command.

Fixed an issue where databases with names longer than 5 characters could not be scaled out after upgrading from Standard Edition to Enterprise Edition.

Fixed an issue where route errors occurred when primary key columns were truncated during automatic sharding.

Fixed an issue where plancache was not cleared after view changes.

Fixed an issue with incorrect diagnostic strategies for index diagnosis on key partition Global Index.

Category

Description

Fixed issues

Fixed an issue with excessively long index names in index recommendations.

Fixed a risk of incorrect pushdown when single table sharding was enabled for single table joins.

Fixed an issue with incorrect partition routing due to incorrect calculation results from the CoHash Right partition function.